Scientific Newsletter - May 4, 2001


log in
Examining Radio Signal Spikes Using Clickplots
SETI@home scientists face the challenge of identifying possible extraterrestrial communication from the enormous number of signals (or "spikes") captured from our radio telescope data. SETI@home users detect about a million strong signals during any given 15 hour period of radio telescope data. As reported in Newsletter #6, most (if not all) of these signals we detect are attributable to noise or RFI (Radio Frequency Interference—signals originating from Earth). Trying to discriminate ET signals is a quintessential needle-in-the-haystack problem.

An important approach to interpreting signals is displaying data in such a way that scientists can visually distinguish unusual signals from the crowd. Unusual signals might have excessively strong signal strength ("power"), demonstrate consistency across frequency, time, and/or location, or occur in unusual patterns that stick out from the field. Waterfall plots (graphs plotting signals across frequency and time) are excellent tools to use for these identification purposes. (See Newsletter #6 for more discussion of waterfall plots.)

Figure 1 below is a waterfall plot of signal spikes found in data from September 23, 2000. The y-axis (time) represents a data tape length of about 57,000 seconds, and the x-axis (frequency) represents the frequency range our data covers (1418.5-1421.5 MHz). Believe it or not, approximately 100,000 workunits were crunched to produce this data, meaning that up to 100,000 SETI@home users contributed to this analysis. These users detected and returned a total of 985,599 spikes. Figure 1 actually plots about 1/16th of these points, since plotting more points would make the graph ridiculously crowded. Since the plotting area in Figure 1 is approximately 220 pixels high by 510 pixels wide, the graph can display a total of 112,200 pixels (220 * 510 = 112,200.) Thus, for readability purposes we want the maximum density to be approximately 50%, or around 61,100 points plotted.

The color coding for signal strengths is as follows:

  • Power 0-20 -> Dark faded blue
  • Power 20-25 -> Dark faded red
  • Power 25-50 -> Red
  • Power 50-100 -> Orange
  • Power 100-1,000 -> Yellow
  • Power >1,000 -> White

As you can see in Figure 1, most signals have strengths between 20-25, with a significant portion also between 25 and 50.

Common Features of a Signal Spike Distribution

Note the features visible in Figure 1. The two vertical yellow lines at frequencies of 1419 and 1421 MHz are "birdies"—test signals injected into the telescope receiver by SETI@home to make sure that the instrumentation and software are working properly. If we didn't see these birdies, we would know that there was a problem with our equipment or coding.

The vertical, thick red/yellow band at 1420 MHz is an artifact of our splitting technology. Without going into excruciating detail, here is an attempt at a brief explanation. When we split frequency data into workunits small enough to be crunched by SETI@home clients, we perform a calculation called a Fast Fourier Transform. A side-effect of this calculation is that signal strengths at the exact center of the frequency band are artificially amplified; the exact center of the frequency band happens to be 1420 MHz. We then remove this bad data, but by doing so we actually lower the signal strength baseline just adjacent to 1420 MHz, which in turn makes "normal" signals with those frequencies appear stronger than they really are (i.e., stronger than the "average" represented by the signal strength baseline). Thus, the band at 1420 MHz looks thicker and smudged because it actually consists of signals just before and just after 1420 MHz.

If you look carefully, you can see some narrow black bands in the plot. These bands represent brief time periods when SETI@home was unable to collect data from the radio telescope. Since we share telescope time with other projects, at times these other projects will quickly refocus the telescope to a different location in the sky. When this movement is particularly quick, we won't be able to accurately collect data, hence producing the black bands you see in Figure 1.

Also, examine the vertical line at about 1420.8 MHz. This is an example of a strong signal at a specific frequency that is consistent across time. It is likely to be some kind of communication broadcast from Earth or from a satellite; further explorations will need to be made to determine its source.

Taking a Closer Look

Now that we have a general idea of how the spikes are distributed across frequency and time, let's take a closer look at some of the more interesting features. Figure 2 below plots spikes within a narrower time and frequency range (between 15,000 and 30,000 seconds and between 1420.6 and 1421.2 MHz, respectively). This plot provides a close-up of both the birdie at 1421 MHz and the vertical band at 1420.8 Mhz. The black horizontal bands are also clear. Also notice the group of white signals at 21,000 seconds and 1421.1 MHz. Let's take an even closer look at that cluster.

Figure 3 below is a waterfall for ranges of 18,750 to 22,500 seconds and 1421.09 to 1421.25 MHz. The white cluster is actually in the shape of a small rectangle about .01 MHz wide and 110 seconds high. This size corresponds to exactly one workunit. (SETI@home users analyze data in increments of one workunit at a time.) Clearly there was a problem with the results returned from this workunit.

Clickplots - Putting It All Together

As demonstrated by the 3 sample plots above, we clearly want the ability to zoom in on particular plot features while also being able to view the big picture. Such a scenario affords great efficiency in identifying and breaking down signal features. We accomplish this task in a straightforward manner—we generate "clickplots". Clickplots are simply GIF files image-mapped so that a mouse click on a particular section of the plot brings up a higher resolution graph of that location. These clickplots are fast, and their production is easily automated, implementable, and maintainable for large quantities of large datasets. In the near future we plan to develop a repository of clickplots viewable by everyone on the web, including plots of gaussians, pulses, and triplets. We also plan to link these plots to a starmap, such that for any point in the sky you can view a clickplot of the signal data collected by SETI@home for that location. Stay tuned and check back for more additions and features.

The following links display clickplots for different categorizations of data obtained on September 23, 2000 (including specific chirp rates and FFT lengths).

  • Complete Dataset
  • Chirp Rate = 0
  • 0 <= Chirp Rate <= 1
  • FFT Length = 128K

    Simple Instructions on How to Use a Clickplot

    Try playing with the above clickplots. Clicking on a particular portion of the plot will load an expanded plot of that portion. Our current plots are clickable 2 levels deep; after two clicks, you'll reach the bottom level of the clickplot. To re-expand, simply hit the "Back" button of your browser.

    Return to Science Newsletter main menu

  • Copyright © 2014 University of California